TDA7293+JFE2140(LSK489)composite feedback amplifier

That was fast!
I see that the THD has improved slightly since the previous picture, but the SNR has dropped by 1dB!

You've obviously hit the "sweet spot" of the THD/SNR balance with the settings and results from @Dxvideos post#113 and your post#114,
which shows a result of 1Wrms/8R0 THD of 0.00093% and a SNR of 98.35dB

In my opinion, this is the OPTIMUM you can expect from this configuration/topology.
The results are very good now, indeed.

And, Now it's time to listen to the AMP prototype for a long time ...and get some impressions.

Best regards
Dragan
 
"Thanks for your support!

In my tests with 4.1mA, 4.6mA, and 5.6mA bias currents, the most pronounced second harmonic was observed at 5.6mA. After various optimizations, I found that the best feedback resistor values are 49.9kΩ and 4.99kΩ, yielding an SNR of 98dB and THD of 0.00115%—a significant improvement.

A slightly stronger second harmonic could contribute to a more musical and pleasing sound. I’ll be evaluating the amplifier with 5.6mA bias and 49.9kΩ/4.99kΩ feedback resistors in listening tests.

5.6mA 49,9k 4,99k graphic
5,6ma normal dirençler   49.9K 4.99K 10x gain  JFE2140 bjt ccs 1w se.jpg
 
I’m currently reviewing your last four THD distribution graphs,
and what stands out to me is the SNR result in the top-right corner of the graph in the last graph, your post ago!

What algorithm does QuantAsylum (QA) use to calculate SNR and THD%?

The 1 kHz carrier signal is 2.839 Vrms, which corresponds to +9.06 dBv (where 1V = 0 dBv).
  • Harmonic II: -92 dBv
  • Harmonic III: -104 dBv
  • Harmonic IV: -97 dBv
  • Harmonic V: -103 dBv
  • Harmonic VI: -106 dBv,
and so on. Higher harmonics are also present, but the lower ones don’t contribute much to the final calculation.

Below the 1 kHz carrier signal, we also observe 50 Hz network interference and its higher harmonics.
This could be either direct interference or forced interference due to lowish PSRR.
  • 50 Hz: -90 dBv
  • 100 Hz: -108 dBv
  • 150 Hz: -102 dBv
  • 200 Hz: -110 dBv
  • 250 Hz: -108 dBv,
and others below -112 dBv, which also don’t contribute much.

The noise floor is seen at approximately <-130 dBv, which is quite uniform across the full bandwidth of 20 Hz to 20 kHz.
There is no increased noise at high frequencies from the TDA7293 output itself, as can be seen from the graph!
Everything is below the mentioned <130dBv !?

When we calculate the distortion sum (higher harmonics + 50 Hz hum) in correlation with the carrier signal at +9 dBv, we get a sum of 96.8 dBv.

If we exclude the lower frequencies below 1 kHz, then the sum is 98.1 dBv (I considered only the harmonics mentioned above), which QA also shows!
But this is effectively the THD(+N) in dB(v), which equates to about 0.00126%.

If we only consider the noise floor for the SNR and include all the interference from the home network and others,
then the displayed SNR is around 98.1 dBv.

However, this doesn’t represent our true picture of the SNR!

What would AQ show if you managed to shield the 50 Hz hum and its harmonics out of equation?

Have you measured the output noise voltage again with the input shorted?
@Dxvideo mentioned measuring around 18 µVrms of broadband output noise, which would give an SNR of around 104 dB @1Wrms@8R0 load.
This is almost the maximum expected for the TDA7293 used.
And it's a very good result, actually!

Cheers, Dragan
 
Last edited:
Hello everyone;
We are approaching the final stage, and a significant improvement in SNR has been achieved, now reaching 103-102 dB. The dominant harmonic remains the second, with noise measuring at 17 µV and DC offset at 0.02 mV. The CCS is set to 4.7 mA.

@Dxvideo has begun critical listening tests and will share their impressions soon.
My initial observations: an exceptionally wide soundstage, pristine highs, and a well-balanced, highly detailed presentation.”
 

Attachments

  • 27C38A0B-6143-491A-BEAC-DEA4A2193907.jpeg
    27C38A0B-6143-491A-BEAC-DEA4A2193907.jpeg
    223.3 KB · Views: 75
My first impressions about the new version;
It has more detailed and more clinical sound than the old version. Wider stage and absolutely no audible noise or distortion at my listenning level!
Basses and mids are still balanced but sounded a bit bright to me.. (must be my personal opinion not an objective assesment)
Sercan has already drawn the PCBs.. The final circuit will be;
Note : R15 should be 4K99 and R13-R14 = 240R..
Adsız.png
 
I am still listening the amplifier.
As I mentioned before; sounded a bit bright to me.
So I thought that putting an R+C filter on the input of the amplifier might solve the problem.
Then tried some resistor and capacitor values like 1K + 1nF, 1K + 4n7 and 1K + 3n3..
Finally I decided that (1K series resistor + 3n3 FKP type parallel capacitor into 1M input load resistor) is a good combination for my ears.
Upper ends not affected in musical meanigs. (in the simulation we loose about 0,6 dB @20Khz)
However mids came forward more.
Details or coloration also not affected. But the music come more relaxing now..
My speakers are Wharfedale Diamond 9.6s.. Recapped with cheap solid caps..
I am 56 years old. So I'm not expecting to hear like 17-18khzs, but this configuration makes me more happy 🙂
Sercan already add a R+C input filter option on the PCB..
I think I can recommend this amplifier to you DIYers..
Note: The phase response bended about -20° at 20Khz in the simulation. But I can't say I felt any auditory impact.
 
Last edited:
Finally, EPIC S Rev2.0 is ready. All details have been updated in post #1. The PCBs have been sent for production. Happy DIY!
 

Attachments

  • FF916CD6-F51F-42D0-9DE7-8423955D5FC4.png
    FF916CD6-F51F-42D0-9DE7-8423955D5FC4.png
    326.2 KB · Views: 163
  • 19D2023D-E48A-47BA-9A25-7174A4204F04.jpeg
    19D2023D-E48A-47BA-9A25-7174A4204F04.jpeg
    296.1 KB · Views: 151
  • 7942FC7A-52B6-4A83-8686-7F9D640BC0DF.jpeg
    7942FC7A-52B6-4A83-8686-7F9D640BC0DF.jpeg
    327.9 KB · Views: 106
  • C4F0F39E-B0C3-44B2-AFDC-9D5D8DC8D2BE.jpeg
    C4F0F39E-B0C3-44B2-AFDC-9D5D8DC8D2BE.jpeg
    504 KB · Views: 105
  • C73FC9B5-7A2D-4CE7-B468-8262CD528128.png
    C73FC9B5-7A2D-4CE7-B468-8262CD528128.png
    993.9 KB · Views: 154
Years ago I made a famous LM3886 for my friend. After designing the Epic S, I asked him to compare the two amplifiers. He said he would continue with the EPIC S and completed the assembly.
 

Attachments

  • 8D9C97D8-CCC5-4477-8622-E226F3BD6E0A.jpeg
    8D9C97D8-CCC5-4477-8622-E226F3BD6E0A.jpeg
    284.2 KB · Views: 85
  • D5903716-51F7-499A-8D4A-9E87E693281A.jpeg
    D5903716-51F7-499A-8D4A-9E87E693281A.jpeg
    386.3 KB · Views: 89
  • 3423E0E7-71B4-43DE-B483-71A54BD71E61.jpeg
    3423E0E7-71B4-43DE-B483-71A54BD71E61.jpeg
    308.2 KB · Views: 87
PCB Available…
I have prepared BOM for all PCB.
EPIC S has two kinds of BOM. One of with CMF series resistors, another is standard metal film resistors.
Enjoy.
 

Attachments

  • EA45BDBA-D684-4353-ACAD-3748AD6B0ADD.jpeg
    EA45BDBA-D684-4353-ACAD-3748AD6B0ADD.jpeg
    680.1 KB · Views: 80
  • D520800E-56CB-490F-928E-3C107DF13A6C.jpeg
    D520800E-56CB-490F-928E-3C107DF13A6C.jpeg
    829.8 KB · Views: 75
  • 18A78857-C672-42C3-809B-2C6A7AA411FA.jpeg
    18A78857-C672-42C3-809B-2C6A7AA411FA.jpeg
    371.3 KB · Views: 59
  • 57078CFF-1776-4443-B112-7B3D88800B8E.jpeg
    57078CFF-1776-4443-B112-7B3D88800B8E.jpeg
    460.5 KB · Views: 67
  • C52911C7-4F72-4F8B-AA9D-ED8D7B17C6AB.jpeg
    C52911C7-4F72-4F8B-AA9D-ED8D7B17C6AB.jpeg
    666.5 KB · Views: 83
EPIC S amplifier.
 

Attachments

  • D639E46D-D9A2-43C3-87A3-C60C5CE349BB.jpeg
    D639E46D-D9A2-43C3-87A3-C60C5CE349BB.jpeg
    554.4 KB · Views: 51
  • 5AE1D19D-1EBF-4455-BBD1-338861B8168A.jpeg
    5AE1D19D-1EBF-4455-BBD1-338861B8168A.jpeg
    433.8 KB · Views: 54
  • ABB3176C-DF8E-4B0E-9430-B0F61E7A08FB.jpeg
    ABB3176C-DF8E-4B0E-9430-B0F61E7A08FB.jpeg
    569.4 KB · Views: 54
  • 8A511C4E-568C-416A-9FD3-C6CA8967C968.jpeg
    8A511C4E-568C-416A-9FD3-C6CA8967C968.jpeg
    658.6 KB · Views: 53
  • ECF8E610-BA36-4FCB-8087-82CE5A6C3388.jpeg
    ECF8E610-BA36-4FCB-8087-82CE5A6C3388.jpeg
    280.9 KB · Views: 49
Nice,
But can't see the point of a such complicated design when the performance is just slightly improved in comparison to the conventional implementation. Theoretically is interesting but in practice not much benefit to the ears. Sorry for the honest opinion.
 
  • Like
Reactions: sercan85